Re: RFC - GL in scheme (long, with code)

[Robert Bruce Findler <robby@cs.rice.edu>]

>  > An alternative implementation would have the C functions assume their
>>  arguments are perfect.  Scheme wrappers would check type and count.
>>  Is there any advantage or disadvantage to this approach?
>
>I recommend checking in C. For the forseeable future, the cost of
>checking is negligible (i.e., relative to other PLT Scheme overheads)
>--- and since the cost is negligible, there's no reason to skip it.

I'm curious -- why wouldn't one rather have private C-based functions 
called things like unsafe:car and provide safe wrapper functions, 
implemented in Scheme like this:

    (define (car x)
      (unless (cons? x) (error 'car ...))
      (unsafe:car x))

I would think that the error code is much easier to implement in 
Scheme than in C, because a few macro abstractions can really 
simplify that checking. In addition, this implementation strategy 
means that optimizations can be implemented as Scheme-based code 
transformations (probably more relevant for things like `car' than 
the ql primitives, tho).

Robby